Abstract
PURPOSE: No formal guidance exists regarding optimal opportunistic computed tomography (CT) region of interest (ROI) size or placement to clinically obtain bone Hounsfield unit (HU) data. Using clinical CT scans, this study evaluated ROI size/placement and assessed HU reproducibility. METHODS: Three non-radiologists independently identified the L1 and L4 vertebral body centroid and then placed varying size circular ROIs on axial and sagittal images of 30 clinical CT scans. A 200-mm(2) ROI location was varied left to right, anterior to posterior, and cranial to caudal. Intra- and inter-observer reliability was determined using intraclass correlation coefficients (ICC). ROI size and axial/sagittal HU comparison was performed by ANOVA and t-test. Precision of 200-mm(2) axial and sagittal ROIs was assessed in a second cohort of 30 patients with two scans obtained within 16 days. RESULTS: Vertebral body centroid placement was nearly identical between readers (ICC > 0.99). Intra- and inter-observer reliability was excellent for all ROI sizes on both projections (ICC > 0.95). Statistically, but not clinically, significant differences, less than 8 HU, were present between various sized ROIs at L1, with no difference at L4. Axial HU was generally higher than sagittal for all ROI sizes at L1 and L4 by ~ 5-12 HU. In the precision cohort, L1 and L4 HU %CV was 4.8-7.9% yielding least significant change values of 10-16 HU. CONCLUSIONS: Non-radiologists can reliably identify the vertebral body centroid and measure HU. For clinical use, we recommend a 200-mm(2) circular ROI placed at the vertebral body centroid on L1 axial imaging.